JP7514813B2 - Sales fraud detection business support device, sales fraud detection business support method, and sales fraud detection business support program - Google Patents

Description

The present invention relates to a sales fraud detection business support device, a sales fraud detection business support method, and a sales fraud detection business support program.

Patent Document 1 discloses a POS terminal and a POS management system that, when it becomes necessary to cancel sales data during sales at the POS terminal, a newly provided misentry request key is pressed to set a misentry request flag in the newly provided sales data, setting the sales as a misentry sale and settling the account, and also displays a list of sales data for which the misentry request flag is set so that the manager can approve and reflect the data, thereby realizing the cancellation of sales with simple operations and displaying and approving a list of cancellation data to prevent fraud (see paragraph 0006 of Patent Document 1).

JP 2004-30210 A

In this way, in fields that aim to detect fraud in sales records (sales fraud), an indicator called the sales cancellation rate is often used. The sales cancellation rate is calculated, for example, by the formula: sales cancellation amount / sales amount. If the sales cancellation rate is a large value, it is possible that fraud such as falsification of sales amounts is taking place, so it is important to carefully check the increase or decrease in the sales cancellation rate.

However, to calculate the sales cancellation rate, the person in charge needs to collect the necessary data by referring to the sales recording history data, etc. This creates the problem that the task of discovering sales fraud takes time and effort.

The present invention has been made in consideration of the above problems, and aims to provide a sales fraud detection support device, a sales fraud detection support method, and a sales fraud detection support program that can support the work of detecting sales fraud by automatically calculating the sales cancellation rate and detecting abnormally increasing values in the calculated sales cancellation rate.

In order to solve the above-mentioned problems and achieve the object, the sales fraud detection business support device of the present invention is a sales fraud detection business support device that has a control unit and a memory unit and can support the business of detecting fraud regarding sales results, the memory unit stores sales amount/sales cancellation data including a period, at least one aggregation unit from among business establishment identification data, department identification data, and person in charge identification data, and at least one aggregation target from among a combination of sales amount and sales cancellation amount and a combination of number of sales and number of sales cancellations, and the control unit, when an amount is specified as the aggregation target for each combination of the period and the aggregation unit, stores the sales amount/sales cancellation data. The system is characterized by comprising a sales cancellation rate calculation means for calculating a sales cancellation rate for each combination of the period and the aggregation unit by dividing the cancellation amount by the sales amount in the sales amount and sales cancellation data, and, when a number of times is specified as the target of aggregation, by dividing the number of sales cancellations in the sales amount and sales cancellation data by the number of sales in the sales amount and sales cancellation data; an abnormal value detection means for detecting, according to a predetermined statistical method, a sales cancellation rate calculated by the sales cancellation rate calculation means that exceeds an upper limit as an abnormal value; and an abnormality display means for displaying the sales cancellation rate detected as an abnormal value by the abnormal value detection means and the combination of the period and the aggregation unit corresponding to the detected sales cancellation rate.

In addition, in the sales fraud detection business support device according to the present invention, the storage unit further stores sales recording history data including records having the period, the counting unit, and sales recording amounts consisting of both positive and negative values, and the control unit further includes an amount counting means for counting the sales recording amounts in the sales recording history data for each combination of the period and the specified counting unit, and the amount counting means calculates the sales cancellation amount for each combination of the period and the specified counting unit by counting the absolute value of the negative amount if the sales recording amount in the newly registered record is a negative amount, and the absolute value of the negative amount if the total value of the sales recording amounts of the two records before and after the correction is a negative amount, and the absolute value of the negative sales recording amount in the deleted record, and calculates the sales amount for each combination of the period and the specified counting unit by counting the sales recording amounts in the records that were not included in the counting of the sales cancellation amount.

In addition, in the sales fraud detection business support device according to the present invention, the storage unit further stores sales recording history data including records having the period, the tallying unit, and sales recording amounts consisting of both positive and negative values, and the control unit further includes a frequency tallying means for tallying the records in the sales recording history data as a number of times for each combination of the period and the specified tallying unit, and the frequency tallying means counts one record as one time when the sales recording amount in the newly registered record is a negative amount, and counts the sales recordings of two records that are in a relationship before and after the correction. If the total of the upper amounts is a negative amount, the two records are counted as one occurrence, and each deleted record is counted as one occurrence to calculate the number of sales cancellations for each combination of the period and the specified aggregation unit, and if the sales posting amount in the newly registered record is a positive amount, the record is counted as one occurrence, and if the total of the sales posting amounts of the two records before and after the correction is a positive amount, the two records are counted as one occurrence to calculate the number of sales cancellations for each combination of the period and the specified aggregation unit.

In addition, in the sales fraud detection business support device according to the present invention, the anomaly display means also displays the specified statistical method and the upper limit value.

The sales fraud detection business support device according to the present invention is also characterized in that the predetermined statistical method is a method using the interquartile range, and the upper limit value is a value calculated by adding the third quartile to a value 1.5 times the interquartile range.

In addition, in the sales fraud detection business support device according to the present invention, the storage unit further stores sales cancellation rate calculation result data including the period, at least one of the aggregation units among the business establishment identification data, the department identification data, and the person in charge identification data, and at least one of the aggregation targets among the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations, and the control unit performs a pre-run calculation for the aggregation unit corresponding to the sales cancellation rate detected by the abnormal value detection means based on the sales cancellation rate calculation result data. The device further includes a period-amount/number-of-times bar graph generating means for generating a bar graph showing the sales amount and the sales cancellation amount for each period, or a bar graph showing the number of sales and the number of sales cancellations for each period for the aggregation unit corresponding to the sales cancellation rate detected by the abnormal value detection means, and the bar graph has the sales amount and the sales cancellation amount in the sales cancellation rate calculation result data, or the number of sales and the number of sales cancellations in the sales cancellation rate calculation result data, on the vertical axis, and the period in the sales cancellation rate calculation result data on the horizontal axis.

In the sales fraud detection business support device according to the present invention, the storage unit further stores sales cancellation rate calculation result data including the period, at least one of the counting units among the business establishment identification data, the department identification data, and the person in charge identification data, and the sales cancellation rate calculated by the sales cancellation rate calculation means, and the control unit further includes period-by-period sales cancellation rate line graph generation means for generating a line graph showing the progress of the sales cancellation rate for each period based on the sales cancellation rate calculation result data for the number of types included in the specific counting unit, and the line graph has the sales cancellation rate in the sales cancellation rate calculation result data as the vertical axis and the period in the sales cancellation rate calculation result data as the horizontal axis.

In addition, in the sales fraud detection business support device according to the present invention, the storage unit further stores sales cancellation rate calculation result data including the period, at least one of the aggregation units among the business establishment identification data, the department identification data, and the person in charge identification data, and at least one of the aggregation targets among the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations, and the control unit calculates the aggregation unit for the period corresponding to the sales cancellation rate detected by the abnormal value detection means based on the sales cancellation rate calculation result data. The device further includes a unit-of-amount/number-of-sales bar graph generating means for generating a bar graph showing the sales amount and the sales cancellation amount for each unit, or a bar graph showing the number of sales and the number of sales cancellations for each unit of the calculation for the period corresponding to the sales cancellation rate detected by the abnormal value detection means, and the bar graph has the sales amount and the sales cancellation amount in the sales cancellation rate calculation result data, or the number of sales and the number of sales cancellations in the sales cancellation rate calculation result data, on the vertical axis, and the unit of the calculation in the sales cancellation rate calculation result data on the horizontal axis.

In addition, in the sales fraud detection business support device according to the present invention, the storage unit further stores daily sales cancellation data including the period, the registration processing date, at least one of the aggregation units among the business establishment identification data, the department identification data, and the person in charge identification data, and at least one of the aggregation targets among the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations, and the control unit generates, based on the daily sales cancellation data, a line graph showing the sales amount for each registration processing date for the combination of the period and the aggregation unit corresponding to the sales cancellation rate detected by the abnormal value detection means, and a line graph showing the sales amount for each registration processing date for the combination of the period and the aggregation unit corresponding to the sales cancellation rate detected by the abnormal value detection means. The system further includes a daily amount/number of times line graph generating means for generating a combination of two graphs, a line graph showing the amount of sales cancellations for each of the registration processing dates for the period and the tally unit combination corresponding to the sales cancellation rate detected by the abnormal value detection means, and a line graph showing the number of sales cancellations for each of the registration processing dates for the period and the tally unit combination corresponding to the sales cancellation rate detected by the abnormal value detection means, and the line graph has the vertical axis representing the sales amount, the sales cancellation amount, the number of sales, or the number of sales cancellations in the daily sales cancellation data, and the horizontal axis representing the registration processing date in the daily sales cancellation data.

In addition, the sales fraud detection business support method according to the present invention is a sales fraud detection business support method that can support the business of detecting fraud related to sales results, which is executed by an information processing device having a control unit and a memory unit, and the memory unit stores sales amount/sales cancellation data including a period, at least one aggregation unit from among business establishment identification data, department identification data, and person in charge identification data, and at least one aggregation target from among a combination of sales amount and sales cancellation amount and a combination of number of sales and number of sales cancellations, and when an amount is specified as the aggregation target for each combination of the period and the aggregation unit executed by the control unit, the sales cancellation amount in the sales amount/sales cancellation data is The method includes a sales cancellation rate calculation step for calculating a sales cancellation rate for each combination of the period and the aggregation unit by dividing the number of sales cancellations in the sales amount and sales cancellation data by the sales amount in the sales amount and sales cancellation data when a number of times is specified as the target of aggregation, an abnormal value detection step for detecting, according to a predetermined statistical method, a sales cancellation rate that exceeds an upper limit value among the sales cancellation rates calculated in the sales cancellation rate calculation step as an abnormal value, and an abnormality display step for displaying the sales cancellation rate detected as an abnormal value in the abnormal value detection step and the combination of the period and the aggregation unit corresponding to the detected sales cancellation rate.

In addition, the sales fraud detection business support program of the present invention is a sales fraud detection business support program that can support the business of detecting fraud regarding sales results, to be executed by an information processing device having a control unit and a memory unit, and the memory unit stores sales amount/sales cancellation data including a period, at least one aggregation unit from among business establishment identification data, department identification data, and person in charge identification data, and at least one aggregation target from among a combination of sales amount and sales cancellation amount and a combination of number of sales and number of sales cancellations, and when an amount is specified as the aggregation target for each combination of the period and the aggregation unit, the sales amount/sales cancellation amount in the sales amount/sales cancellation data is stored. The method includes a sales cancellation rate calculation step of calculating a sales cancellation rate for each combination of the period and the aggregation unit by dividing the cancellation amount by the sales amount in the sales amount and sales cancellation data, and when a number of times is specified as the target of aggregation, by dividing the number of sales cancellations in the sales amount and sales cancellation data by the number of sales in the sales amount and sales cancellation data; an abnormal value detection step of detecting, according to a predetermined statistical method, a sales cancellation rate calculated in the sales cancellation rate calculation step that exceeds an upper limit as an abnormal value; and an abnormality display step of displaying the sales cancellation rate detected as an abnormal value in the abnormal value detection step and the combination of the period and the aggregation unit corresponding to the detected sales cancellation rate.

The present invention has the effect of supporting the detection of sales fraud by automatically calculating the sales cancellation rate and detecting abnormally increasing values in the calculated sales cancellation rate.

FIG. 1 is a block diagram showing an example of the configuration of a sales fraud detection operation support device. FIG. 2 is a diagram showing an example of fraudulent sales records. FIG. 3 is a diagram showing an example of an image of calculation of the sales cancellation rate. FIG. 4 is a diagram showing an example of a sales cancellation amount. FIG. 5 is a diagram showing an example of the initial analysis screen. FIG. 6 is a diagram showing an example of the analysis screen. FIG. 7 is a diagram showing an example of sales data to be corrected and accumulated sales posting history data. FIG. 8 is a diagram showing an example of a sales data registration screen and sales entry history data in normal sales data registration (Pattern 1). FIG. 9 is a diagram showing an example of a sales data registration screen and sales recording history data when an increase correction is made to sales data (Pattern 2). FIG. 10 is a diagram showing an example of a sales data registration screen and sales recording history data when a reduction correction is registered for sales data (Pattern 3). FIG. 11 is a diagram showing an example of a sales data registration screen and sales recording history data when deleting and registering sales data (Pattern 4). FIG. 12 is a diagram showing an example of a sales data registration screen and sales entry history data when negative sales data (red slip) is registered (Pattern 5). FIG. 13 is a diagram showing an example of automatic detection execution schedule data and parameters used for acquiring the data. FIG. 14 is a diagram showing an example of the acquisition range condition data and parameters used for acquiring the data. FIG. 15 is a diagram showing an example of parameters used for acquiring data from the sales cancellation rate calculation result data. FIG. 16 is a diagram showing an example of sales recording history data, sales amount/sales cancellation data, and sales cancellation rate calculation result data. FIG. 17 is a diagram showing an example of abnormal value detection using the interquartile range. FIG. 18 is a diagram showing an example of the abnormality determination result data, the abnormality determination result message data, and the abnormality determination result message detail data. FIG. 19 is a diagram showing an example of parameters used to acquire the abnormality determination result message data. FIG. 20 is a diagram showing an example of a summary message displayed on the initial analysis screen. FIG. 21 is a diagram showing an example of a set of reference dates on the initial analysis screen. FIG. 22 is a diagram showing an example of a detailed message displayed on the analysis screen. FIG. 23 is a diagram showing an example of an output area of a graph. FIG. 24 is a diagram showing an example of analysis data acquisition range conditions. FIG. 25 is a diagram showing an example of extraction condition initial values. FIG. 26 is a diagram showing an example of a set of extraction conditions on the analysis screen. FIG. 27 is a diagram showing an example in which the calculation unit "person in charge" and the calculation target "accrued amount" are specified as parameters used to acquire data that is the basis for generating graphs 1 to 3. FIG. 28 is a diagram showing an example of sales recording history data, sales amount/sales cancellation data, and sales cancellation rate calculation result data. FIG. 29 is a diagram showing an example in which the aggregation unit "person in charge" and the aggregation target "accrued amount" are specified as parameters used to acquire data that is the basis for generating graph 4. FIG. 30 is a diagram showing an example of sales recording history data and daily sales cancellation data. FIG. 31 is a diagram showing an example of graph 1 in the case where the aggregation unit is "person in charge" and the aggregation target is "accrued amount." FIG. 32 is a diagram showing an example of graph 2 in the case where the aggregation unit is "person in charge" and the aggregation target is "accrued amount." FIG. 33 is a diagram showing an example of graph 3 in the case where the aggregation unit is "person in charge" and the aggregation target is "accrued amount." FIG. 34 is a diagram showing an example of graph 4 in the case where the tally unit is "person in charge" and the tally target is "accrued amount." FIG. 35 is a diagram showing an example of a list of graphs 1 to 4 when the aggregation unit is "person in charge" and the aggregation target is "accrued amount." FIG. 36 is a diagram showing an example of coloring parameters that need to be set when highlighting a graph by coloring. FIG. 37 is a diagram showing an example in which the calculation unit "person in charge" and the calculation target "number of times recorded" are specified as parameters used to acquire data that is the basis for generating graphs 1 to 3. FIG. 38 is a diagram showing an example of sales recording history data, sales amount/sales cancellation data, and sales cancellation rate calculation result data. FIG. 39 is a diagram showing an example in which the tally unit "person in charge" and the tally target "number of times recorded" are specified as parameters used to acquire data that is the basis for generating graph 4. FIG. 40 is a diagram showing an example of sales recording history data and daily sales cancellation data. FIG. 41 is a diagram showing an example of graph 1 in the case where the counting unit is "person in charge" and the counting target is "number of times recorded." FIG. 42 is a diagram showing an example of graph 2 in the case where the counting unit is "person in charge" and the counting target is "number of times recorded." FIG. 43 is a diagram showing an example of graph 3 in the case where the counting unit is "person in charge" and the counting target is "number of times recorded." FIG. 44 is a diagram showing an example of graph 4 in the case where the counting unit is "person in charge" and the counting target is "number of times recorded." FIG. 45 is a diagram showing an example of a list of graphs 1 to 4 when the tally unit is "person in charge" and the tally target is "number of times recorded." FIG. 46 is a diagram showing an example of coloring parameters that need to be set when highlighting a graph by coloring. FIG. 47 is a diagram showing an example of a comparison of graph 3 when the aggregation target is the "accrued amount" and when the aggregation target is the "accrued number of times." FIG. 48 is a diagram showing an example in which the aggregation unit "department" and the aggregation target "accrued amount" are specified as parameters used to acquire the data that is the basis for generating graphs 1 to 3. FIG. 49 is a diagram showing an example of sales recording history data, sales amount/sales cancellation data, and sales cancellation rate calculation result data. FIG. 50 is a diagram showing an example in which the aggregation unit "department" and the aggregation target "accrued amount" are specified as parameters used to acquire data that is the basis for generating graph 4. FIG. 51 is a diagram showing an example of sales recording history data and daily sales cancellation data. FIG. 52 is a diagram showing an example of graph 1 when the aggregation unit is "department" and the aggregation target is "accrued amount." FIG. 53 is a diagram showing an example of graph 2 when the aggregation unit is "department" and the aggregation target is "accrued amount." FIG. 54 is a diagram showing an example of graph 3 when the aggregation unit is "department" and the aggregation target is "accrued amount." FIG. 55 is a diagram showing an example of graph 4 when the aggregation unit is "department" and the aggregation target is "accrued amount." FIG. 56 is a diagram showing an example of a list of graphs 1 to 4 when the aggregation unit is "department" and the aggregation target is "accrued amount." FIG. 57 is a diagram showing an example of coloring parameters that need to be set when highlighting a graph by coloring. FIG. 58 is a diagram showing an example of a comparison of graph 2 when the aggregation unit is "person in charge" and when the aggregation unit is "department."

Below, an embodiment of a sales fraud detection business support device, a sales fraud detection business support method, and a sales fraud detection business support program according to the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to the present embodiment.

[1. Overview]
In this section, the background, problems, overview and effects of the present invention will be explained in order.

[1-1. Background]
In recent years, corporate fraud and scandals such as embezzlement by employees and executives (hereinafter collectively referred to as "fraud") have been on the rise. In addition, due to the impact of COVID-19, such fraud is on the rise. As a result, fraud is exceeding the range of what can be detected by humans, and there is a demand for a system that can detect and deal with it early using non-human means.

There are various types of fraud. The first type is fraud related to sales records. This type of fraud inflates sales figures by falsifying sales amounts. The second type is fraud related to purchasing records. This type of fraud involves collusion with business partners through fictitious payments. The third type is fraud related to inventory. This type of fraud involves profit manipulation, such as by inflating end-of-period inventory values. This embodiment deals with the first of these types of fraud, fraud related to sales records (hereinafter referred to as "sales fraud").

Sales fraud is usually committed in order to meet sales targets for each employee or organization. Sales fraud is particularly prevalent in subsidiaries, and parent companies are required to take measures to prevent fraud.

One way to commit fraudulent sales is to increase the amount of sales by a large amount at the end of a period and then cancel the large amount at the beginning of the following period, as shown in Figure 2.

When sales fraud occurs, the amount of sales cancellations generated will be clearly high compared to the actual sales figures when viewed in monthly sales results. For this reason, as shown in Figure 3, if the sales cancellation rate is calculated from the sales amount and sales cancellation amount within a single month, it is possible to detect as an abnormality a person or organization with an abnormally high calculated sales cancellation rate.

When focusing on monetary amounts, the sales cancellation rate can be calculated, for example, using the formula "sales cancellation amount/sales amount." The sales cancellation amount in this formula is an absolute value. Conventionally, sales fraud was discovered by personnel manually calculating the sales cancellation rate by referencing vast amounts of data.

[1-2. Issues]
However, there are problems with the detection of sales fraud by humans, such as the following (1) and (2).

(1) First, the first problem was that it took a lot of time to tally up sales records and cancellation records for each person in charge or organization. In other words, in order to discover fraudulent sales, it was necessary to target all people in charge or organizations, tally up all data on sales records and sales recording history for each target, extract and tally up only data that had been reduced from sales in order to check cancellation records, and then compare and confirm the tallied sales records and sales cancellation records. In short, because it was necessary to narrow down conditions from a huge amount of transaction data and check the tally, this was a time-consuming task and involved a high risk of mistakes.

(2) The second problem is that it is necessary to compare and confirm the actual results at the end of the term and the beginning of the following term, where fraud is likely to occur, for each person in charge or for each organization, which takes time and effort, and there is a risk of mistakes or omissions in confirmation. In other words, in order to discover sales fraud, it is necessary to check the actual results at the end of the previous term and the beginning of the following term, compare the sales actuals and cancellation actuals between these two, and check for unnatural sales actuals and unnatural sales cancellations that occur from month to month. In short, since it is often not possible to determine whether sales fraud has occurred by checking only the data for a single month, it is only at the end of the following term when the results for multiple months are checked that sales fraud is discovered, which causes the problem of it taking a long time to discover fraud.

[1-3. Overview and Effects]
Therefore, in this embodiment, in order to solve the above problems (1) and (2), for example, the following functions (1) and (2) are implemented.

(1) First, the sales cancellation rate is calculated from the sales amount and sales cancellation amount within a single month, making it possible to automatically detect abnormally high sales cancellation rates. Specifically, the sales recording history for a single month is referenced to automatically calculate the sales cancellation rate for each person in charge or each organization, and the user can be notified of months and person in charge or organization in which the sales cancellation rate is abnormally high compared to the months throughout the year. This makes it possible, for example, to successfully eliminate the cost of aggregating sales performance for each person in charge or each organization.

(2) In addition, by visualizing and comparing the sales amount and the amount of sales cancellation for the month in which an anomaly was detected and the surrounding months, it is now possible to show the basis for determining that the amount of sales cancellation for the month in which an anomaly was detected is abnormal. Specifically, by making it possible to check the breakdown of the sales cancellation rate that is notified when an anomaly is determined, it is now possible to analyze the validity of the amount of sales cancellation based on the size of the amount and the actual occurrence trend from previous months. In short, it is now possible to check the two axes of sales amount and amount of sales cancellation simultaneously at once when comparing multiple months.

In summary, in this embodiment, for example, fraudulent patterns are organized into scenarios and automatically detected periodically, making it possible to detect sales fraud early and deal with it automatically rather than manually.

In the present specification and drawings, the term "sales cancellation amount" refers to the amount deducted from actual sales. Specifically, as shown in the right column of the table in Figure 4, the "sales cancellation amount" refers to the amount deducted when a negative amount is recorded in sales, the amount deducted when a sales amount is revised downward, and the amount deleted when sales are deleted. The specific configuration and operation are described below.

2. Configuration
An example of the configuration of the sales fraud detection operation support device 100 according to this embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of the configuration of the sales fraud detection operation support device 100.

The sales fraud detection business support device 100 is a commercially available desktop personal computer. Note that the sales fraud detection business support device 100 is not limited to a stationary information processing device such as a desktop personal computer, but may be a portable information processing device such as a commercially available notebook personal computer, PDA (Personal Digital Assistant), smartphone, or tablet personal computer.

The sales fraud detection business support device 100 comprises a control unit 102, a communication interface unit 104, a memory unit 106, and an input/output interface unit 108. Each unit of the sales fraud detection business support device 100 is connected to each other so that they can communicate with each other via any communication path.

The communication interface unit 104 communicatively connects the sales fraud detection business support device 100 to the network 300 via a communication device such as a router and a wired or wireless communication line such as a dedicated line. The communication interface unit 104 has a function of communicating data with other devices via the communication line. Here, the network 300 has a function of connecting the sales fraud detection business support device 100 and the server 200 so that they can communicate with each other, and is, for example, the Internet or a LAN (Local Area Network). Note that data such as various masters described below may be stored in the server 200, for example.

An input device 112 and an output device 114 are connected to the input/output interface unit 108. The output device 114 may be a monitor (including a home television), a speaker, or a printer. The input device 112 may be a keyboard, a mouse, a microphone, or a monitor that works with a mouse to provide a pointing device function. In the following, the output device 114 may be referred to as the monitor 114, and the input device 112 may be referred to as the keyboard 112 or the mouse 112.

The storage unit 106 stores various databases, tables, files, and the like. The storage unit 106 records computer programs that work in cooperation with the OS (Operating System) to give instructions to the CPU (Central Processing Unit) to perform various processes. The storage unit 106 can be, for example, a memory device such as a RAM (Random Access Memory) or a ROM (Read Only Memory), a fixed disk device such as a hard disk, a flexible disk, an optical disk, or the like.

The memory unit 106 includes, for example, automatic detection execution schedule data 106a, acquisition range condition data 106b, sales recording history data 106c, sales amount/sales cancellation data 106d, sales cancellation rate calculation result data 106e, daily sales cancellation data 106f, anomaly determination result data 106g, anomaly determination result message data 106h, and anomaly determination result message detail data 106i.

Before going into detail about the contents of each piece of data, we will explain the overview and position of each piece of data.

The automatic detection execution schedule data 106a and the acquisition range condition data 106b belong to the "data for performing anomaly detection" and are pre-configured data for performing anomaly detection.

Sales recording history data 106c, sales amount/sales cancellation data 106d, sales cancellation rate calculation result data 106e, and daily sales cancellation data 106f belong to "business data" and are assumed data accumulated during business operations. Sales recording history data 106c includes data on sales recording, sales correction, and sales deletion performed during business operations. The history of corrections consists of red invoice data before correction and black invoice data after correction. Figure 7 shows sales recording history data 106c accumulated when a sales amount of 1,000 yen recorded on 4/15/2021 is corrected to 700 yen on 4/20/2021.

The abnormality determination result data 106g, the abnormality determination result message data 106h, and the abnormality determination result message detail data 106i belong to the "abnormality determination result data" and are data that store the determination results of the abnormality detection execution.

The contents of each piece of data are explained in detail below.

As shown in FIG. 13, the automatic detection execution schedule data 106a includes, for example, a detection ID, a schedule ID, execution conditions, and execution time.

The acquisition range condition data 106b is, for example, data for setting the range when acquiring data from the sales recording history data 106c. As shown in FIG. 14, the acquisition range condition data 106b includes, for example, the detection ID, the schedule ID, a target column, a FROM condition, a TO condition, and the like.

In this embodiment, at least one of the combinations of the sales amount and the sales cancellation amount, and the combination of the number of sales and the number of sales cancellations is the target of aggregation. Also, in this embodiment, at least one of the business establishment identification data (business establishment), department identification data (department), and person in charge identification data (person in charge) is the unit of aggregation when aggregating the target of aggregation.

The sales recording history data 106c is data that accumulates daily sales recording. As shown in Figures 28, 30, 38, 40, 49, and 51, the sales recording history data 106c includes records having, for example, a registration processing classification, a history serial number, a period (accounting year and month), a slip number, a sales date, the business establishment identification data (business establishment), the department identification data (department), the person in charge identification data (person in charge), a sales recording amount (sales amount), and a registration processing date. The sales recording amount (sales amount) consists of both positive and negative values. Examples of the registration processing classification include a classification indicating that the record is newly registered (new registration), a classification indicating that the record is before correction (before correction), a classification indicating that the record is after correction (after correction), and a classification indicating that the record is deleted (deletion).

Here, we will explain five patterns for accumulating business sales recording history data 106c.

As shown in Figure 8, the first pattern is normal sales data registration. The details indicated by the history serial number "1" in the sales recording history data 106c in Figure 8 are the details registered in the first pattern.

The second pattern is the registration of an increase in sales data, as shown in Figure 9. Related history is linked by the document number. In addition, for correction data, if one detail has a registration process classification of "before correction" and the other detail has a registration process classification of "after correction", and the history serial numbers of these two details are consecutive, these two details are considered as one set. In the example of sales recording history data 106c in Figure 9, a detail with a registration process classification of "before correction" and history serial number "2" and a detail with a registration process classification of "after correction" and history serial number "3" form one set. The two details indicated by history serial numbers "2" and "3" in the sales recording history data 106c in Figure 9 are the details registered in the second pattern.

The third pattern is the registration of a reduction correction to sales data, as shown in Figure 10. The idea of linking voucher numbers and the idea of grouping details before and after correction into one set are the same as in the second pattern. The two details indicated by the history sequence numbers "4" and "5" in the sales recording history data 106c in Figure 10 are the details registered in the third pattern.

As a fourth pattern, as shown in FIG. 11, there is a deletion and registration of sales data. The concept of linking the invoice numbers is the same as in the second pattern. The details indicated by the history sequence number "6" in the sales recording history data 106c in FIG. 11 are the details registered in the fourth pattern.

As a fifth pattern, as shown in FIG. 12, there is the registration of negative sales data (red invoices). The detail indicated by the history sequence number "7" in the sales recording history data 106c in FIG. 12 is the detail registered in the fifth pattern.

Returning to the explanation of the data contained in the memory unit 106, the sales amount/sales cancellation data 106d is data that includes the sales amount and the sales cancellation amount, or the number of sales and the number of sales cancellations, calculated by referring to the sales recording history data 106c, for each combination of the period (fiscal year/month) and the specified aggregation unit. As shown in Figures 28, 38, and 49, the sales amount/sales cancellation data 106d includes, for example, the period (fiscal year/month), at least one aggregation unit among the business establishment identification data (business establishment), the department identification data (department), and the person in charge identification data (person in charge), and at least one aggregation target among the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations, etc.

The sales cancellation rate calculation result data 106e is data that includes the sales cancellation rate calculated with reference to the sales amount/sales cancellation data 106d for each combination of the period (fiscal year/month) and the specified aggregation unit. As shown in Figures 28, 38, and 49, the sales cancellation rate calculation result data 106e includes, for example, the period (fiscal year/month), at least one aggregation unit from among the business establishment identification data (business establishment), the department identification data (department), and the person in charge identification data (person in charge), at least one aggregation target from among the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations, and the calculated sales cancellation rate.

The sales cancellation rate calculation result data 106e may further include, for example, a negative correction amount, a deletion amount, and a negative sales amount as a breakdown of the sales cancellation amount. The sales cancellation rate calculation result data 106e may further include, for example, the number of negative corrections, the number of deletions, and the number of negative sales as a breakdown of the number of sales cancellations.

Daily sales cancellation data 106f is data including the sales amount and the sales cancellation amount, or the number of sales and the number of sales cancellations, calculated with reference to sales recording history data 106c, for each combination of the period (fiscal year and month), the registration processing date, and the specified aggregation unit. As shown in Figures 30, 40, and 51, daily sales cancellation data 106f includes, for example, the period (fiscal year and month), the registration processing date, at least one aggregation unit among the business establishment identification data (business establishment), the department identification data (department), and the person in charge identification data (person in charge), and at least one aggregation target among the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations, etc.

As shown in FIG. 18 etc., the abnormality determination result data 106g includes, for example, an abnormality detection ID (detection ID) that is the detection ID that detected an abnormality in the sales cancellation rate (hereinafter referred to as "anomaly detection" in this paragraph), an abnormality detection JOB ID (JOBID) that is an ID for identifying the JOB in which the abnormality was detected, a message ID that is an ID for identifying the message to be output when the abnormality is detected, the period (fiscal year and month) in which the abnormality was detected, the establishment identification data (establishment) for the establishment in which the abnormality was detected, the department identification data (department) for the department in which the abnormality was detected, the person in charge identification data (person in charge) for the person in charge who detected the abnormality, the detected abnormal sales cancellation rate, and an upper limit value calculated using a specified statistical method.

As shown in FIG. 18 etc., the abnormality determination result message data 106h includes, for example, the abnormality detection ID (detection ID), the abnormality detection JOBID (JOBID), the message ID, a category (anomaly level) for identifying the presence or absence of an abnormality, a definition name specified by the abnormality detection JOBID, an overview, the detected content (detection target), etc. The detected content is the detected abnormal sales cancellation rate and the corresponding combination of the period (fiscal year and month) and the person in charge identification data (person in charge).

As shown in FIG. 18, etc., the abnormality determination result message detail data 106i includes, for example, the abnormality detection ID (detection ID), the abnormality detection JOBID (JOBID), the message ID, the specified statistical method (detection method), a threshold value, a determination method, the upper limit value, etc.

The control unit 102 is a CPU or the like that provides overall control of the sales fraud detection business support device 100. The control unit 102 has an internal memory for storing control programs such as an OS, programs that define various processing procedures, and required data, and executes various information processing operations based on these stored programs.

The control unit 102, in terms of its functional concept, includes, for example, (1) an amount counting unit 102a as an amount counting means for counting the sales recorded amounts in the sales recorded history data for each combination of the period and the specified counting unit, (2) a number counting unit 102b as a number counting means for counting the records in the sales recorded history data as a number of times for each combination of the period and the specified counting unit, and (3) a sales cancellation unit that calculates a sales cancellation rate for each combination of the period and the counting unit by dividing the sales cancellation amount in the sales amount/sales cancellation data by the sales amount in the sales amount/sales cancellation data when an amount is specified as the target of counting for each combination of the period and the counting unit, and by dividing the number of sales cancellations in the sales amount/sales cancellation data by the number of sales in the sales amount/sales cancellation data when a number is specified as the target of counting. (4) an abnormal value detection unit 102d as an abnormal value detection means for detecting, according to a predetermined statistical method, a sales cancellation rate calculated by the sales cancellation rate calculation means that exceeds an upper limit as an abnormal value; (5) an abnormality display unit 102e as an abnormality display means for displaying the sales cancellation rate detected as an abnormal value by the abnormal value detection means and the combination of the period and the counting unit corresponding to the detected sales cancellation rate; and (6) a period-amount/number of times bar graph generation means for generating, based on the sales cancellation rate calculation result data, a bar graph showing the sales amount and the sales cancellation amount for each period for the counting unit corresponding to the sales cancellation rate detected by the abnormal value detection means, or a bar graph showing the number of sales and the number of sales cancellations for each period for the counting unit corresponding to the sales cancellation rate detected by the abnormal value detection means. (7) a period-sales cancellation rate line graph generating unit 102g as a period-sales cancellation rate line graph generating means which generates a line graph showing a trend of the sales cancellation rate for each period for the number of types included in the specific aggregation unit based on the sales cancellation rate calculation result data; (8) a counting unit-amount/number bar graph generating unit 102h as a counting unit-amount/number bar graph generating means which generates a bar graph showing the sales amount and the sales cancellation amount for each aggregation unit for the period corresponding to the sales cancellation rate detected by the abnormal value detection means, or a bar graph showing the number of sales and the number of sales cancellations for each aggregation unit for the period corresponding to the sales cancellation rate detected by the abnormal value detection means, based on the sales cancellation rate calculation result data; and a daily-amount/number-of-times line graph generating unit 102i as daily-amount/number-of-times line graph generating means for generating a combination of two graphs, a line graph showing the sales amount for each registration processing date for the combination of the period and the tally unit corresponding to the sales cancellation rate detected by the abnormal value detection means, and a line graph showing the sales cancellation amount for each registration processing date for the combination of the period and the tally unit corresponding to the sales cancellation rate detected by the abnormal value detection means, or a combination of two graphs, a line graph showing the number of sales for each registration processing date for the combination of the period and the tally unit corresponding to the sales cancellation rate detected by the abnormal value detection means, and a line graph showing the number of sales cancellations for each registration processing date for the combination of the period and the tally unit corresponding to the sales cancellation rate detected by the abnormal value detection means.

The amount tallying unit 102a tally up the sales recorded amounts in the sales recorded history data 106c, which includes the records having the fiscal year/month, the tallying unit, and the sales recorded amounts that are both positive and negative values, for each combination of the fiscal year/month and the specified tallying unit.

Specifically, the amount tallying unit 102a calculates the sales cancellation amount for each combination of the fiscal year/month and the specified tallying unit by tallying the absolute value of the negative amount if the sales posting amount in the newly registered record is a negative amount, the absolute value of the negative amount if the sum of the sales posting amounts of the two records before and after the correction is a negative amount, and the absolute value of the negative sales posting amount in the deleted record. In addition, the amount tallying unit 102a calculates the sales amount for each combination of the fiscal year/month and the specified tallying unit by tallying the sales posting amounts in the records that were not included in the tally of the sales cancellation amount.

The frequency counting unit 102b counts, for each combination of the accounting year and month and the specified counting unit, the records in the sales recording history data 106c that include the records having the accounting year and month, the counting unit, and the sales recording amount that is both positive and negative.

Specifically, the number of times counting unit 102b calculates the number of sales cancellations for each combination of the accounting year/month and the specified counting unit by counting one record for each newly registered record if the sales posting amount in the newly registered record is a negative amount, counting the two records for each deleted record if the total value of the sales posting amounts of the two records before and after the correction is a negative amount, and counting one deleted record for each deleted record. Also, the number of times counting unit 102b calculates the number of sales cancellations for each combination of the accounting year/month and the specified counting unit by counting one record for each newly registered record if the sales posting amount in the newly registered record is a positive amount, and counting the two records for each deleted record if the total value of the sales posting amounts of the two records before and after the correction is a positive amount.

The sales cancellation rate calculation unit 102c calculates the sales cancellation rate for each combination of the fiscal year/month and the aggregation unit based on the sales amount/sales cancellation data 106d, which includes the fiscal year/month, the aggregation unit, and at least one aggregation target of the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations, by dividing the sales cancellation amount in the sales amount/sales cancellation data 106d by the sales amount in the sales amount/sales cancellation data 106d when "amount" is specified as the aggregation target for each combination of the fiscal year/month and the aggregation unit. In addition, when "number of times" is specified as the target of aggregation for each combination of the fiscal year/month and the aggregation unit based on the sales amount/sales cancellation data 106d, the sales cancellation rate calculation unit 102c calculates the sales cancellation rate for each combination of the fiscal year/month and the aggregation unit by dividing the number of sales cancellations in the sales amount/sales cancellation data 106d by the number of sales in the sales amount/sales cancellation data 106d.

The abnormal value detection unit 102d detects, according to a predetermined statistical method, any of the sales cancellation rates calculated by the sales cancellation rate calculation unit 102c that exceed an upper limit as an abnormal value. The predetermined statistical method is not particularly limited, but is, for example, a method using the interquartile range. When the interquartile range is used, the upper limit is, for example, a value calculated by adding a value obtained by multiplying the interquartile range by 1.5 to the third quartile.

The abnormality display unit 102e displays the sales cancellation rate detected as an abnormal value by the abnormal value detection unit 102d and the combination of the accounting year/month and the aggregation unit corresponding to the detected sales cancellation rate. At this time, the abnormality display unit 102e may also display the specified statistical method and the upper limit value.

The period-amount/number bar graph generating unit 102f generates a bar graph (see Figures 31 and 52) showing the sales amount and the sales cancellation amount for each fiscal year and month for the aggregation unit corresponding to the sales cancellation rate detected by the abnormal value detection unit 102d, or a bar graph (see Figure 41) showing the number of sales and the number of sales cancellations for each fiscal year and month for the aggregation unit corresponding to the sales cancellation rate detected by the abnormal value detection unit 102d, based on the sales cancellation rate calculation result data 106e, which includes the fiscal year and month, the aggregation unit, and at least one of the aggregation targets of the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations.

The vertical and horizontal axes of the bar graph described in the previous paragraph are as follows:
Vertical axis: the sales amount and the sales cancellation amount in the sales cancellation rate calculation result data 106e, or the number of sales and the number of sales cancellations in the sales cancellation rate calculation result data 106e. Horizontal axis: the accounting year and month in the sales cancellation rate calculation result data 106e.

The sales cancellation rate by period line graph generating unit 102g generates line graphs (see Figures 32, 42, and 53) showing the progress of the sales cancellation rate for each fiscal year and month, for the number of types included in the specific aggregation unit, based on the sales cancellation rate calculation result data 106e, which includes the fiscal year and month, the aggregation unit, and the sales cancellation rate calculated by the sales cancellation rate calculation unit 102c.

The vertical and horizontal axes of the line graph described in the previous paragraph are as follows:
Vertical axis: the sales cancellation rate in the sales cancellation rate calculation result data 106e Horizontal axis: the accounting year and month in the sales cancellation rate calculation result data 106e

The amount/number of sales bar graph generating unit 102h generates a bar graph (see Figures 33 and 54) showing the sales amount and the sales cancellation amount for each aggregation unit for the fiscal year/month corresponding to the sales cancellation rate detected by the abnormal value detection unit 102d, or a bar graph (see Figure 43) showing the number of sales and the number of sales cancellations for each aggregation unit for the fiscal year/month corresponding to the sales cancellation rate detected by the abnormal value detection unit 102d, based on the sales cancellation rate calculation result data 106e, which includes the fiscal year/month, the aggregation unit, and at least one of the aggregation targets of the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations.

The vertical and horizontal axes of the bar graph described in the previous paragraph are as follows:
Vertical axis: the sales amount and the sales cancellation amount in the sales cancellation rate calculation result data 106e, or the number of sales and the number of sales cancellations in the sales cancellation rate calculation result data 106e. Horizontal axis: the aggregation unit in the sales cancellation rate calculation result data 106e.

The daily-amount/number-of-times line graph generating unit 102i generates a line graph showing the sales amount for each registration processing date for the combination of the fiscal year/month and the aggregation unit corresponding to the sales cancellation rate detected by the anomaly detection unit 102d, and a line graph showing the sales amount for each registration processing date for the combination of the fiscal year/month and the aggregation unit corresponding to the sales cancellation rate detected by the anomaly detection unit 102d, based on the daily sales cancellation data 106f, which includes the fiscal year/month, the registration processing date, the aggregation unit, and at least one of the aggregation targets of the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations. A combination of two graphs is generated: a line graph showing the amount of sales cancellations for each registration processing date for a combination of counting units (see Figures 34 and 55); or a combination of two graphs is generated: a line graph showing the number of sales for each registration processing date for a combination of the fiscal year and month and the counting units corresponding to the sales cancellation rate detected by the abnormal value detection unit 102d, and a line graph showing the number of sales cancellations for each registration processing date for a combination of the fiscal year and month and the counting units corresponding to the sales cancellation rate detected by the abnormal value detection unit 102d (see Figure 44).

The vertical and horizontal axes of the line graph described in the previous paragraph are as follows:
Vertical axis: the sales amount, the sales cancellation amount, the number of sales, or the number of sales cancellations in the daily sales cancellation data 106f. Horizontal axis: the registration processing date in the daily sales cancellation data 106f.

[3. Processing Overview and Screen Configuration]
[3-1. Processing Overview]
In this embodiment, the "sales cancellation rate against actual sales" is output. Furthermore, in this embodiment, "business data related to the detected information" is also output. Information for which an abnormality is detected may be highlighted by changing the color or font size so that the abnormality can be recognized at a glance. The data to be detected is, for example, the sales cancellation rate against actual sales for each combination of month and person in charge. For example, a person in charge with an abnormally large amount of sales cancellations in a single month is detected.

[3-2. screen structure]
In this embodiment, the displayed screen changes in the following order: (1)→(2).

(1) Display of anomaly detection results First, as shown by A in the example screen of Fig. 5, the results of detecting an anomaly in the sales cancellation rate are displayed at the summary message level. The area shown by A in Fig. 5 is an area for displaying messages related to the anomaly detection process. Specifically, this area displays the timing at which the anomaly was detected, the person in charge, and the sales cancellation rate. Note that, if there are multiple pieces of information detected as an anomaly, the multiple pieces of information are displayed vertically, as shown by A in Fig. 5. This area only displays the anomaly detection results on a summary basis, so if you want to see the anomaly detection results on a detailed basis, you need to select the summary message and switch screens.

(2) Display of business data related to the detected abnormal information Next, detailed messages and graphs necessary for analyzing the detected abnormal information are displayed.

The area shown in the lower half of A in Figure 6 is an area that displays detailed messages regarding the anomaly detection results. This area displays information such as the "detection method used to detect the anomaly" and "key information of the data in which the anomaly was detected." If multiple pieces of information are detected as an anomaly, a summary message will be displayed for each anomaly, but a detailed message will only be displayed for the one selected from the displayed summary messages. Furthermore, after the detailed message is displayed, the following graphs B to E are displayed.

The graph shown in area B of Figure 6 is a graph showing the actual sales amount and the amount of sales cancellations recorded by a specific person in charge. In other words, it is a graph for checking "how much actual sales amount and amount of sales cancellations have occurred each month for a person in whom an abnormality was detected." Note that the person to be output can be switched by specifying it, so it is also possible to check the difference in trends with other people in charge.

The graph shown in area C of Figure 6 is a graph showing the sales cancellation rate for each employee. In other words, it is a graph for checking "the trend in sales cancellation rate against actual sales amount for each month and employee." This graph makes it possible to compare the trend in sales cancellation rate for the employee in whom an abnormality was detected with other employees.

The graph shown in area D of Figure 6 is a graph showing the actual sales amount and the amount of sales cancellations for each employee within a single month. In other words, it is a graph for confirming the "breakdown of the actual sales amount and the amount of sales cancellations for each employee." This graph makes it possible to compare the difference in the amount of corrections incurred for the employee in whom an abnormality was detected and other employees. Note that cancellation methods include posting a negative amount, deletion, and reduction correction, and this graph can also display the breakdown of each cancellation method.

The graph shown in area E of Figure 6 is a graph showing the sales amount and sales cancellation amount by day within a single month, recorded by a specific person in charge. In other words, it is a graph for checking "the sales amount by date and the sales cancellation amount by date recorded by a person in whom an abnormality was detected." This graph makes it possible to check at what point in a single month actual results were recorded, whether there are any periods of concentration, etc.

Area F in Figure 6 is an area for setting the extraction conditions for data extraction. In other words, this area is used when you want to display the results of data extraction with different conditions for the graphs displayed in areas B to E in Figure 6. For example, if you want to switch and analyze output data at the organizational level, you would specify "Establishment" or "Department" in area F in Figure 6. Also, if you want to switch and analyze from the perspective of the number of times sales are registered, you would specify "Number of times recorded" in area F in Figure 6.

4. Specific examples of processing
In this section, a specific example of the processing according to this embodiment will be described. In [4-1], the processing (anomaly detection processing) of automatically detecting a person in charge with an abnormally high sales cancellation rate relative to the actual sales by checking the contents of the sales recording history data 106c in the business data will be described. In [4-2], the processing (analysis screen display processing) of displaying the detected abnormal data and data related to the abnormal data that was referenced for the detection on the analysis screen will be described.

[4-1: Abnormality detection process]
(1) Pre-setting As the pre-setting, information required for anomaly detection is stored in advance in a table. In this example, it is assumed that the automatic detection execution schedule data 106a is set and stored as shown in Fig. 13, and the acquisition range condition data 106b is set and stored as shown in Fig. 14.

(2) Automatic Execution (2-1) Acquisition of Automatic Detection Execution Schedule Data 106a First, the automatic detection execution schedule data 106a set in (1) is automatically acquired. Specifically, if "detection ID: AB001, schedule ID: SH001" is set as parameters as shown in Fig. 13, the automatic detection execution schedule data 106a shown in Fig. 13 is automatically acquired as data having the detection ID and schedule ID.

(2-2) Determining whether it is the timing to detect an anomaly Next, it is determined whether the timing automatically acquired in (2-1) is the timing to detect an anomaly. Specifically, it is assumed that the timing automatically acquired in (2-1) is "2021/04/07 (the fifth business day of April)." It is possible to determine holidays and business days by referencing the calendar master in the business database, and it is assumed that the calendar master is always updated based on the sales business calendar.

Now, referring to the automatic detection execution schedule data 106a in FIG. 13 that was automatically acquired in (2-1), the execution condition is "the fifth business day of every month." In this way, if the timing at which the automatic detection execution schedule data 106a was automatically acquired matches the timing defined by the execution condition set in the automatic detection execution schedule data 106a, the execution decision result will be "execute," and the processing from (2-3) onwards will be executed. On the other hand, if there is no match, the execution decision result will be "do not execute," and the processing will end.

(2-3) Acquisition of range conditions of data for detecting anomalies Next, the acquisition range condition data 106b set in (1) is automatically acquired. Specifically, the acquisition range condition data 106b shown in FIG. 14 is automatically acquired as data having the parameters "detection ID: AB001, schedule ID: SH001" set in (2-1).

(2-4) Detection of Personnel with Abnormally High Sales Cancellation Rate Finally, by referring to the sales recording history data 106c in the business data, personnel with an abnormally high sales cancellation rate are detected. In this example, the sales recording history data 106c is set and saved as shown in FIG.

(2-4-1) Calculation of Sales Cancellation Rate First, the amount tallying unit 102a refers to the sales recording history data 106c in FIG. 16, and tallies the sales cancellation amounts and sales amounts based on the following rules.
(rule)
The sales cancellation amount is calculated by totaling the absolute value of the amounts in the records that meet any of the following totaling conditions a to c.
a. The registration process classification is "new registration" and the amount is negative b. The registration process classification is "correction" and the total amount before and after correction is negative c. The registration process classification is "deletion"
Regarding the sales amount, the amounts in the records that do not meet the aggregation conditions for the sales cancellation amount are aggregated.

Assume that the parameters "Extraction condition column: accounting year/month, FROM condition: 2020/05, TO condition: 2021/04" are automatically set as shown in FIG. 15 based on the "Sales recording history data acquisition range condition data 106b" shown in FIG. 14 that was automatically acquired in (2-3). In this case, the amount aggregation unit 102a aggregates only records with accounting years/months 2020/05 to 2021/04 from the sales recording history data 106c in FIG. 16.

The tabulation results are as shown in the sales amount/sales cancellation data 106d in FIG. 16. Next, the sales cancellation rate calculation unit 102c converts the sales cancellation rate into a percentage value based on the sales cancellation rate calculation formula (sales cancellation amount/sales amount) and obtains it. The calculation results are as shown in the sales cancellation rate calculation result data 106e in FIG. 16.

(2-4-2) Detection of Abnormal Values Next, the upper and lower limits are obtained from the sales cancellation rate calculation result data 106e using the "quartile range," which is one of the statistical methods, and abnormal values are detected based on the upper limit. FIG. 17 is a diagram showing how to obtain the upper limit (20%) and the lower limit (0%) when focusing on the person in charge A in the sales cancellation rate calculation result data 106e of FIG. 16. As shown in FIG. 17, when the sales cancellation rate for person in charge A is plotted in a box-and-whisker plot, the sales cancellation rate of 31% in April 2021 exceeds the upper limit of 20%. Therefore, the abnormal value detection unit 102d detects the sales cancellation rate of 31% in April 2021 as an abnormal value. In the example of this paragraph, only the detection of abnormalities for person in charge A has been described for the sake of simplicity, but in actual processing, the abnormal value detection unit 102d detects abnormal values for all persons in charge.

Here, three representative statistical methods will be described. First, standard deviation is a method of taking the distribution of values at a certain point, setting a confidence interval, and judging normality/abnormality. Standard deviation is an analysis that does not include the concept of a time axis, so it is not suitable for cross-sectional data analysis of "tendencies from the past". Secondly, the interquartile range used in this example is a method of determining the analysis range of data, aggregating it on the vertical axis, determining the quartile range, and specifying the range magnification that can be judged as normal. The interquartile range is compatible with cross-sectional data analysis because it treats both time series data and point data as vertical axis elements. Thirdly, moving average + outlier is a method of determining the range of the average value and calculating it for each time series. For moving average + outlier, it is necessary to set a threshold value in advance to judge abnormality from the average value, but it is possible to detect abnormal increases and decreases in the transition of the time series. In this embodiment, the interquartile range and moving average + outlier are raised as candidates, but for moving average + outlier, the user needs to set a threshold value of the average value to judge abnormality. For this reason, in this embodiment, it is preferable to use the interquartile range, which can automatically calculate a threshold and determine abnormality.

(2-4-3) Storage of Abnormal Values Finally, the abnormal value detection unit 102d stores information such as the abnormality determination result (accounting year and month: 2021/04, person in charge: person in charge A, sales cancellation rate: 31%) in a table, and at the same time, stores a message to be displayed on the analysis screen. FIG. 18 shows abnormality determination result data 106g, abnormality determination result message data 106h, and abnormality determination result message detail data 106i generated by storing the abnormality determination result.

[4-2: Analysis screen display process]
(1) Display of summary message Next, information that is automatically detected as an anomaly from the sales recording history data 106c in the business data is output to the initial analysis screen. The acquisition and display of the initial analysis screen will be described in detail below in (1-1) to (1-3).

(1-1) Acquiring a result message of an abnormality detection First, assume that "JOBID: sales cancellation alert" is set as a parameter, as shown in FIG. 19. Note that, in order to display the detected information in a list, the detection ID is set to an unset state, as shown in FIG. 19. In this case, the abnormality display unit 102e acquires the abnormality determination result data 106g, abnormality determination result message data 106h, and abnormality determination result message detail data 106i in FIG. 18, which were generated in (2-4-3), as data having "JOBID: sales cancellation alert." These acquired data are also shown in FIG. 20.

(1-2) Displaying on screen a message resulting from the detection of an anomaly Next, the anomaly display unit 102e displays the degree of anomaly "x", the definition name "Sales cancellation alert", the summary "Person in charge A detected", and the detection target "2021/04 Person in charge A 31%" in the anomaly determination result message data 106h of Fig. 20 acquired in (1-1) on the initial analysis screen shown in Fig. 20. The displayed degree of anomaly, definition name, summary, and detection target are collectively referred to as a summary message.

This allows the operator to know that there was a staff member with an abnormally high sales cancellation rate, and who that staff member was (in this example, Staff A). At this time, the operator can also know the accounting month and year in which the abnormality occurred (in this example, 2021/04), and the sales cancellation rate value that was detected as an abnormal value (in this example, 31%).

(1-3) Setting the reference date After the abnormality is displayed, the date when the initial analysis screen was launched (2021/05/10) is set as the reference date in the extraction condition area on the initial analysis screen, as shown in Figure 21.

(2) Display of detailed message and graphs Next, the detailed message and four graphs are output to the analysis screen. The display of the detailed message and the acquisition and display of the four graphs will be explained below in (2-1) to (2-4).

(2-1) Displaying detailed messages and securing output areas for graphs 1. First, when the summary message displayed in (1-2) is selected, an analysis screen is launched. The anomaly display unit 102e displays the detection method "interquartile range," the threshold "1.5 times normal range," and the upper limit "20%" in the anomaly determination result message detail data 106i acquired in (1-1) on the analysis screen shown in Fig. 22. The displayed detection method, threshold, and upper limit are collectively referred to as a detailed message.

In this way, by displaying the summary message explained in (1-2) and the detailed message explained in this section (2-1), the operator can understand the target (person in charge) for which the abnormality was detected, the detection method (calculation method), the benchmark value (sales cancellation rate) for determining that an abnormality was detected, etc. This allows the operator to perform an analysis while knowing which person has an abnormal sales cancellation rate. The sales cancellation rate of a person with an abnormal sales cancellation rate and the past performance that is the basis for determining that an abnormality was detected can be analyzed using the four graphs explained below.

2. Next, an output area for the graphs is secured. In this embodiment, four graphs are output, so an output area for four graphs is secured, as shown by the four blank spaces on the analysis screen in FIG. 23.

(2-2) Setting Extraction Conditions Next, in order to output a graph, it is necessary to extract the data to be output from the analysis data. Prior to the extraction, however, the extraction conditions are set using the method described in this section (2-2).

1. First, the acquisition range of the analysis data is obtained based on the accounting year and month (2021/04) held in the anomaly determination result data 106g in FIG. 20 that is linked to the summary message displayed in (1-2).

(i) Specifically, the accounting year and month within a range of one year before and after the accounting year and month in which the anomaly was detected (2021/04) is acquired as the acquisition range. In other words, the accounting year and month one year ago is 2020/04, and the accounting year and month one year later is 2022/04.

The setting of acquiring accounting dates within a range of one year before and after is in accordance with the condition of detecting within a one-year range in the "Sales Recording History Data Acquisition Range Condition Data 106b," which is data provided in advance for anomaly detection. If you wish to expand the range for analysis, you can do so by changing the start date and end date of the period in the extraction conditions area on the analysis screen.

(ii) Based on the acquired acquisition range (2020/04 to 2022/04), the analysis data acquisition range condition is set by narrowing down the range in which the analysis data exists. In this example, it is assumed that the analysis data exists only in the range of 2020/05 to 2021/04. In this case, the analysis data acquisition range condition is as follows, as shown in FIG. 24:
・Extraction condition column: Fiscal year and month ・FROM condition: 2020/05 (taken from the month after 2020/04, which is the fiscal year and month one year ago)
・TO condition: 2021/04 (hiring from the month before 2022/04, which is the fiscal year one year later)
It becomes.

2. Next, the analysis data acquisition range conditions set in "1." are set as initial values, which become the extraction condition initial values (base date, period start, and period end). The extraction condition initial values are shown in FIG. 25. These extraction condition initial values are then set in the extraction condition area on the analysis screen, as shown in FIG. 26. Note that the period start date and period end date set in this area can be manually changed by the operator.

Then, when the display state is switched to the graph, the extraction items for extracting data for graph output are displayed. As shown in FIG. 26, the "Accounted amount" is set as the fixed initial value for the aggregation target (display switching), and the "Person in charge" is set as the aggregation unit (aggregation) and displayed. In this example (2), these are assumed to remain as fixed initial values and are not changed.

(2-3) Acquisition of Data to Generate Graphs Next, analysis data is acquired based on the extraction conditions set in (2-2). Hereinafter, the bar graph showing the monthly sales amount (number of sales) and sales cancellation amount (number of sales cancellations) for a specific person will be referred to as "Graph 1." Also, the line graph showing the trend in the monthly sales cancellation rate for each person will be referred to as "Graph 2." And the bar graph showing the sales amount (number of sales) and sales cancellation amount (number of sales cancellations) for each person in a specific month will be referred to as "Graph 3." And the line graph showing the trend in the daily sales cancellation rate for a specific month and a specific person will be referred to as "Graph 4."

The acquisition of the data that is the basis for generating graphs 1 to 3 will be described below in (2-3-1), and the acquisition of the data that is the basis for generating graph 4 will be described below in (2-3-2).

(2-3-1) Acquisition of Data to Generate Graphs 1 to 3 Based on the extraction conditions set in (2-2) (shown again in the example screen of FIG. 27), parameters for acquiring graphs 1 to 3 are set as shown in FIG. 27. In the parameters shown in FIG. 27, the start of the period "2020/05", the end of the period "2021/04", the display mode "accounted amount", and the aggregation mode "person in charge" are set based on the extraction conditions set in (2-2), and the designated accounting year and month "2021/04" is set based on the anomaly determination result data 106g.

Then, based on the set parameters shown in FIG. 27, the sales cancellation rate calculation result data 106e, which is the data from which graphs 1 to 3 are generated, is generated as follows.

That is, the amount tallying unit 102a tallies the sales cancellation amount and the sales amount based on the following rules, with reference to the sales recording history data 106c in Fig. 28. Note that the amount tallying unit 102a only targets records in the sales recording history data 106c in Fig. 28 that have accounting years and months that fall within the period start "2020/05" to period end "2021/04" set by the parameters shown in Fig. 27.
(rule)
The sales cancellation amount is calculated by totaling the absolute value of the amounts in the records that meet any of the following totaling conditions a to c.
a. The registration process classification is "new registration" and the amount is negative b. The registration process classification is "correction" and the total amount before and after correction is negative c. The registration process classification is "deletion"
Regarding the sales amount, the amounts in the records that do not meet the aggregation conditions for the sales cancellation amount are aggregated.

As shown in the parameters of FIG. 27, in this example (2), the display mode is specified as "recorded amount" and the aggregation mode is specified as "person in charge," so the aggregation results include the sales amount and sales cancellation amount for each combination of accounting year/month and person in charge, as shown in the sales amount/sales cancellation data 106d of FIG. 28.

The sales cancellation rate calculation unit 102c calculates the sales cancellation rate for each combination of accounting month and person in charge based on the sales amount/sales cancellation data 106d in FIG. 28. The sales cancellation rate calculation unit 102c converts the sales cancellation rate into a percentage value based on the calculation formula for the sales cancellation rate when the display mode is "recorded amount" (sales cancellation amount/sales amount). Note that the sales cancellation rate calculation unit 102c may include not only the sales cancellation rate for each combination of accounting month and person in charge, but also the sales amount, sales cancellation amount, negative correction amount, deletion amount, and negative sales amount for each combination of accounting month and person in charge, as shown in FIG. 28. The generated sales cancellation rate calculation result data 106e is shown in FIG. 28.

(2-3-2) Acquisition of Data to be Used to Generate Graph 4 Based on the extraction conditions set in (2-2) (shown again in the example screen of FIG. 29), parameters for acquiring Graph 4 are set as shown in FIG. 29. In the parameters shown in FIG. 29, the display mode "Accounted Amount" and the aggregation mode "Person in Charge" are set based on the extraction conditions set in (2-2), and the designated accounting year and month "2021/04" and the designated person in charge "Person in Charge A" are set based on the anomaly determination result data 106g.

Then, based on the set parameters shown in FIG. 29, the daily sales cancellation data 106f, which is the data from which Graph 4 is generated, is generated as follows.

In other words, the amount tallying unit 102a refers to the sales recording history data 106c in FIG. 30 and tallies the sales cancellation amount and sales amount based on the same rules as in (2-3-1). However, in this item (2-3-2), in order to obtain data by day, the sales cancellation amount and sales amount are tallied for each combination of accounting year/month, registration processing date, and person in charge. The tallying results are as shown in the daily sales cancellation data 106f in FIG. 30.

(2-4) Graph output (display)
Next, four graphs are output based on the sales cancellation rate calculation result data 106e or the daily sales cancellation data 106f generated in (2-3).

1. Figure 31 shows the data binding information used when binding data to graph 1 and the graph 1 that is generated. The period-amount/number of times bar graph generation unit 102f generates a bar graph (graph 1 in Figure 31) with the sales amount and sales cancellation amount on the vertical axis and the accounting year/month on the horizontal axis for person A, who was the person in charge with an abnormal sales cancellation rate, based on the sales cancellation rate calculation result data 106e in Figure 28.

By looking at Graph 1, the operator can check the sales amount and sales cancellation amount for the employee for whom an anomaly was detected. In addition, it is also possible to compare by month to see the difference in the occurrence status between the month in which an anomaly was detected and other months. Graph 1 in Figure 31 shows that the amount of sales cancellations in 2021/04 was abnormally high.

2. Figure 32 shows the data binding information used when binding data to graph 2 and the graph 2 that is generated. The sales cancellation rate by period line graph generation unit 102g generates a line graph (graph 2 in Figure 32) with the sales cancellation rate on the vertical axis and the accounting year and month on the horizontal axis for each person in charge, based on the sales cancellation rate calculation result data 106e in Figure 28.

By looking at Graph 2, the operator can check the rate of sales cancellations by representative. In addition, based on the sales cancellation rate calculated from the sales amount and sales cancellation amount confirmed in Graph 1, it is also possible to compare the tendency of sales cancellations to occur by representative. Graph 2 in Figure 32 shows that the sales cancellation rate for Representative A in April 2021 is clearly high.

3. The data binding information used when binding data to Graph 3 and the generated Graph 3 are shown in FIG. 33. The amount/number of times bar graph generation unit by aggregation unit 102h generates a bar graph (Graph 3 in FIG. 33) for 2021/04, with the sales amount and sales cancellation amount (negative correction amount, deletion amount, and negative sales amount) on the vertical axis and the person in charge on the horizontal axis, based on the sales cancellation rate calculation result data 106e in FIG. 28. Note that "2021/04" displayed on Graph 3 is based on the accounting year and month held by the anomaly determination result data 106g.

By looking at Graph 3, the operator can confirm the amount incurred by each sales accounting method for a specific staff member. There are three elements related to sales cancellations: negative correction, deletion, and negative sales. By checking the results for each of these three elements, the operator can confirm whether the staff member is using various cancellation methods to cancel. Compared to using only one method, performing cancellation processing using various cancellation methods can make the cancellation amount per element appear smaller, but by looking at Graph 3, it is possible to discover more malicious cancellation processing that uses various cancellation methods like this.

4. The data binding information used when binding data to graph 4 and the generated graph 4 are shown in FIG. 34. The daily amount/number of times line graph generation unit 102i generates a line graph (graph 4 in FIG. 34) with the sales amount on the vertical axis and the registration processing date on the horizontal axis, and a line graph (graph 4 in FIG. 34) with the sales cancellation amount on the vertical axis and the registration processing date on the horizontal axis, for 2021/04 and person in charge A, based on the daily sales cancellation data 106f in FIG. 30. Note that "2021/04" and "person in charge A" displayed in graph 4 are based on the accounting year/month and person in charge held by the anomaly determination result data 106g, respectively.

By looking at Graph 4, an operator can confirm the trend in the recording of sales and sales cancellations by recording date for a specific month and specific person in charge. In the case of fraud in which fictitious sales are recorded at the end of the period and then deleted early in the beginning of the following period, it is conceivable that a large amount of sales cancellations will occur at the beginning of the month. Graph 4 in Figure 34 is a graph that visualizes this type of fraudulent case. Graph 4 in Figure 34 confirms that a large amount of sales cancellations occurs concentrated at the beginning of the month.

5. The generated graphs 1 to 4 are displayed in a list on the analysis screen, as shown in Figure 35.

6. Data for which an abnormality has been detected may be highlighted (e.g., displayed in a different color) based on the information of the person in charge included in the abnormality determination result data 106g linked to the abnormality determination result message detail data 106i. For example, when displaying graph 2, as shown in FIG. 36, the color of the abnormality flag in the legend indicating true (red) may be applied to the person in charge for whom the abnormality was detected (= the person in charge included in the abnormality determination result data 106g), while the color of the abnormality flag in the legend indicating false (gray) may be applied to the people other than the person in charge for whom the abnormality was detected.

In addition to graph 2, it is also possible to highlight (for example, by changing the color) using a similar method. The information that is highlighted is information related to sales cancellations. That is, in the case of graph 1, the bar graph of "Sales Cancellation Amount" is highlighted. In the case of graph 3, the bar graph's "Negative Correction Amount," "Deletion Amount," and "Above Negative Sales Amount" are highlighted. And in the case of graph 4, the line graph of "Sales Cancellation Amount" is highlighted.

(3) Switching the display of a graph Next, switching the display of an analysis graph is explained in the following (3-1) and (3-2). In the explanation so far, a case where "accounted amount" is specified as the aggregation target (display switching) and "person in charge" is specified as the aggregation unit (aggregation) is explained. In contrast, in the following (3-1), a case where "accounted number of times" is specified as the aggregation target (display switching) and "person in charge" is specified as the aggregation unit (aggregation) is explained. In the following (3-2), a case where "accounted amount" is specified as the aggregation target (display switching) and "department" is specified as the aggregation unit (aggregation) is explained.

(3-1) Display of analysis graph when "recording count" is specified as the aggregation target and "person in charge" is specified as the aggregation unit In this section (3-1), an example of confirming and analyzing the situation based on the number of sales recorded using the display switching function is explained. In this section, data focusing on the sales correction count rate in the number of sales recorded can be output.

1. First, as shown in the example screen in Figure 37, in the area for setting extraction conditions on the analysis screen, "Number of times recorded" is selected as the item to be tallied (display switch). In other words, the item to be tallied (display switch) is changed from "Amount recorded" to "Number of times recorded." Note that the item for the aggregation unit (aggregation) remains "Person in charge," as shown in Figure 37.

Based on the set extraction conditions, parameters for obtaining graphs 1 to 3 are set as shown in FIG. 37. In the parameters shown in FIG. 37, the start of the period "2020/05", the end of the period "2021/04", the display mode "Number of entries", and the aggregation mode "Person in charge" are set based on the extraction conditions described in the previous paragraph, and the specified accounting year and month "2021/04" is set based on the anomaly determination result data 106g.

2. Then, based on the set parameters shown in FIG. 37, the sales cancellation rate calculation result data 106e, which is the data from which graphs 1 to 3 are generated, is generated as follows.

That is, the number of times tallying up unit 102b tallies the number of sales cancellations and the number of sales based on the following rules, with reference to the sales recording history data 106c in Fig. 38. Note that, among the sales recording history data 106c in Fig. 38, the number of times tallying up unit 102b targets only records having accounting years and months that belong to the period start "2020/05" to period end "2021/04" set by the parameters shown in Fig. 37.
(rule)
The number of sales cancellations is calculated by counting records that meet any of the following counting conditions a to c.
a. The registration process classification is "new registration" and the amount is negative b. The registration process classification is "correction" and the total amount before and after correction is negative c. The registration process classification is "deletion"
Regarding the number of sales, records that meet any of the following aggregation conditions d to e are aggregated.
d. The registration processing classification is "new registration" and the amount is positive. e. The registration processing classification is "correction" and the total amount before and after correction is positive.

As shown in the parameters of FIG. 37, in this example (3-1), "Number of times recorded" is specified as the display mode and "Person in charge" is specified as the aggregation mode, so the aggregation results include the number of sales and the number of sales cancellations for each combination of accounting year/month and person in charge, as shown in the sales amount/sales cancellation data 106d of FIG. 38.

The sales cancellation rate calculation unit 102c calculates the sales cancellation rate for each combination of accounting month and person in charge based on the sales amount/sales cancellation data 106d in FIG. 38. The sales cancellation rate calculation unit 102c converts the sales cancellation rate into a percentage value based on the calculation formula for the sales cancellation rate when the display mode is "number of times recorded" (number of sales cancellations/number of sales). Note that the sales cancellation rate calculation unit 102c may include not only the sales cancellation rate for each combination of accounting month and person in charge, but also the number of sales, number of sales cancellations, number of negative corrections, number of deletions, and number of negative sales for each combination of accounting month and person in charge, as shown in FIG. 38. The generated sales cancellation rate calculation result data 106e is shown in FIG. 38.

3. Next, based on the extraction conditions set in "1", parameters for obtaining graph 4 are set as shown in FIG. 39. In the parameters shown in FIG. 39, the display mode "Number of times recorded" and the aggregation mode "Person in charge" are set based on the extraction conditions described in "1.", and the designated accounting year and month "2021/04" and the designated person in charge "Person in charge A" are set based on the anomaly determination result data 106g.

Then, based on the set parameters shown in FIG. 39, the daily sales cancellation data 106f, which is the data from which Graph 4 is generated, is generated as follows.

That is, the number of times counting unit 102b refers to the sales recording history data 106c in FIG. 40 and counts the number of sales cancellations and the number of sales based on the same rules as in "2.". However, in this example "3.", in order to obtain data by day, the number of sales and the number of sales cancellations are counted for each combination of accounting year/month, registration processing date, and person in charge. The counting results are as shown in the daily sales cancellation data 106f in FIG. 40.

4. Next, four graphs are output based on the sales cancellation rate calculation result data 106e generated in "2." or the daily sales cancellation data 106f generated in "3.".

The data binding information used when binding data to Graph 1 and the generated Graph 1 are shown in Figure 41. The period-amount/number of times bar graph generation unit 102f generates a bar graph (Graph 1 in Figure 41) for Person A, who was the person in charge with an abnormal sales cancellation rate, with the number of sales and the number of sales cancellations on the vertical axis and the accounting year and month on the horizontal axis, based on the sales cancellation rate calculation result data 106e in Figure 38.

Figure 42 shows the data binding information used when binding data to graph 2 and the graph 2 that is generated. The sales cancellation rate by period line graph generation unit 102g generates a line graph (graph 2 in Figure 42) for each person in charge, with the sales cancellation rate on the vertical axis and the accounting year and month on the horizontal axis, based on the sales cancellation rate calculation result data 106e in Figure 38.

The data binding information used when binding data to Graph 3 and the generated Graph 3 are shown in FIG. 43. The tabulation unit-amount/number of times bar graph generation unit 102h generates a bar graph (Graph 3 in FIG. 43) for 2021/04, with the number of sales and the number of sales cancellations (number of negative corrections, number of deletions, and number of negative sales) on the vertical axis and the person in charge on the horizontal axis, based on the sales cancellation rate calculation result data 106e in FIG. 38. Note that "2021/04" displayed on Graph 3 is based on the accounting year and month held by the anomaly determination result data 106g.

The data binding information used when binding data to graph 4 and the generated graph 4 are shown in FIG. 44. The daily amount/number of times line graph generation unit 102i generates a line graph (graph 4 in FIG. 44) with the number of sales on the vertical axis and the registration processing date on the horizontal axis, and a line graph (graph 4 in FIG. 44) with the number of sales cancellations on the vertical axis and the registration processing date on the horizontal axis, for 2021/04 and person in charge A, based on the daily sales cancellation data 106f in FIG. 40. Note that "2021/04" and "person in charge A" displayed in graph 4 are based on the accounting year/month and person in charge held in the anomaly determination result data 106g, respectively.

5. The generated graphs 1 to 4 are displayed in a list on the analysis screen, as shown in Figure 45.

6. Data for which an abnormality has been detected may be highlighted (e.g., displayed in a different color) based on the information of the person in charge included in the abnormality determination result data 106g linked to the abnormality determination result message detail data 106i. For example, when displaying graph 2, as shown in FIG. 46, the color (red) of the abnormality flag in the legend may be applied to the person in charge for whom the abnormality was detected (= the person in charge included in the abnormality determination result data 106g), and the color (gray) of the abnormality flag in the legend may be applied to the person other than the person in charge for whom the abnormality was detected.

In addition to graph 2, it is also possible to highlight (for example, by changing the color) using a similar method. The information related to sales cancellations is highlighted. That is, in the case of graph 1, the bar graph of "number of sales cancellations" is highlighted. In addition, in the case of graph 3, the bar graph of "number of negative corrections," "number of deletions," and "number of negative sales" is highlighted. And in the case of graph 4, the line graph of "number of sales cancellations" is highlighted.

7. Here, it is important to check the following points regarding the displayed graphs 1 to 4.

First, whether there are a large number of cancellations not only in the sales booked amount but also in the number of sales booked. A large number of cancellations in the number of sales booked means that cancellations are being made little by little across multiple vouchers. In other words, by canceling across multiple vouchers, it is possible that sales are being cancelled in a way that makes it difficult to detect the fraud. In this way, by focusing on the number of sales booked, when conducting a later investigation into the actual fraud, it will be possible to carry out the investigation with the information that multiple small cancellations have occurred across multiple vouchers.

Also, regarding multiple cancellation operations (negative correction, negative sales, and deletion), it is important to consider whether a large number of cancellations have been performed for each.

As a concrete example, FIG. 47 shows a side-by-side comparison of Graph 3 (shown in FIG. 33) generated based on the recorded amount and Graph 3 (shown in FIG. 43) generated based on the recorded number of times. As FIG. 47 shows, it can be seen that Person A performed many cancellation-related operations in terms of both the recorded amount and the recorded number of times, which suggests the possibility that Person A is "intentionally" performing multiple cancellation operations to conceal fraud.

(3-2) Display of analysis graph when "recorded amount" is specified as the aggregation target and "department" is specified as the aggregation unit In this section (3-2), an example of using the display switching function to confirm and analyze the recording status of sales amounts and sales cancellation amounts by organization will be described. In this section (3-2), aggregation is performed by department, and person in charge A included in the anomaly judgment result data 106g is assumed to belong to department A.

1. First, as shown in the example screen in Figure 48, in the area for setting extraction conditions on the analysis screen, "department" is selected as the item for the aggregation unit (aggregation). In other words, the item for the aggregation unit (aggregation) is changed from "person in charge" to "department." Note that the item to be aggregated (display switch) remains "accounted amount," as shown in Figure 48.

Based on the set extraction conditions, parameters for obtaining graphs 1 to 3 are set as shown in FIG. 48. In the parameters shown in FIG. 48, the start of the period "2020/05", the end of the period "2021/04", the display mode "Accounted amount", and the aggregation mode "Department" are set based on the extraction conditions described in the previous paragraph, and the specified accounting year and month "2021/04" is set based on the anomaly determination result data 106g.

2. Then, based on the set parameters shown in FIG. 48, the sales cancellation rate calculation result data 106e, which is the data from which graphs 1 to 3 are generated, is generated as follows.

That is, the amount tallying unit 102a tallies the sales cancellation amount and the sales amount based on the following rules, with reference to the sales recording history data 106c in Fig. 49. Note that the amount tallying unit 102a only targets records in the sales recording history data 106c in Fig. 49 that have accounting years and months that fall within the period start "2020/05" to period end "2021/04" set by the parameters shown in Fig. 48.
(rule)
The sales cancellation amount is calculated by totaling the absolute value of the amounts in the records that meet any of the following totaling conditions a to c.
a. The registration process classification is "new registration" and the amount is negative b. The registration process classification is "correction" and the total amount before and after correction is negative c. The registration process classification is "deletion"
Regarding the sales amount, the amounts in the records that do not meet the aggregation conditions for the sales cancellation amount are aggregated.

As shown in the parameters in FIG. 48, in this example (3-2), the display mode is specified as "recorded amount" and the aggregation mode is specified as "department," so the aggregation results include the sales amount and sales cancellation amount for each combination of accounting year/month and department, as shown in the sales amount/sales cancellation data 106d in FIG. 49.

The sales cancellation rate calculation unit 102c calculates the sales cancellation rate for each combination of accounting month and department based on the sales amount/sales cancellation data 106d in FIG. 49. The sales cancellation rate calculation unit 102c converts the sales cancellation rate into a percentage value based on the calculation formula for the sales cancellation rate when the display mode is "recorded amount" (sales cancellation amount/sales amount). Note that the sales cancellation rate calculation unit 102c may include not only the sales cancellation rate for each combination of accounting month and department, but also the sales amount, sales cancellation amount, negative correction amount, deletion amount, and negative sales amount for each combination of accounting month and department, as shown in FIG. 49. The generated sales cancellation rate calculation result data 106e is shown in FIG. 49.

3. Next, based on the extraction conditions set in "1", parameters for obtaining graph 4 are set as shown in FIG. 50. In the parameters shown in FIG. 50, the display mode "Accounted amount" and the aggregation mode "Department" are set based on the extraction conditions described in "1.", and the designated accounting year and month "2021/04" and the designated department "Department A" are set based on the anomaly determination result data 106g.

Then, based on the set parameters shown in FIG. 50, the daily sales cancellation data 106f, which is the data from which Graph 4 is generated, is generated as follows.

That is, the number counting unit 102b refers to the sales recording history data 106c in FIG. 51 and counts the sales cancellation amount and sales amount based on the same rules as in "2.". However, in this example "3.", in order to obtain data by day, the sales amount and sales cancellation amount are counted for each combination of accounting year/month, registration processing date, and department. The counting results are as shown in the daily sales cancellation data 106f in FIG. 51.

4. Next, four graphs are output based on the sales cancellation rate calculation result data 106e generated in "2." or the daily sales cancellation data 106f generated in "3.".

The data binding information used when binding data to Graph 1 and the generated Graph 1 are shown in Figure 52. The period-amount/number of times bar graph generation unit 102f generates a bar graph (Graph 1 in Figure 52) with the sales amount and sales cancellation amount on the vertical axis and the accounting year/month on the horizontal axis, for "Department A," the department to which Person A, the person in charge with an abnormal sales cancellation rate, belongs, based on the sales cancellation rate calculation result data 106e in Figure 49.

Figure 53 shows the data binding information used when binding data to graph 2 and the graph 2 that is generated. The sales cancellation rate by period line graph generation unit 102g generates a line graph (graph 2 in Figure 53) for each department, with the sales cancellation rate on the vertical axis and the accounting year and month on the horizontal axis, based on the sales cancellation rate calculation result data 106e in Figure 49.

The data binding information used when binding data to graph 3 and the graph 3 that is generated are shown in FIG. 54. The amount/number of times bar graph generation unit by aggregation unit 102h generates a bar graph (graph 3 in FIG. 54) for 2021/04, with the sales amount and sales cancellation amount (negative correction amount, deletion amount, and negative sales amount) on the vertical axis and the department on the horizontal axis, based on the sales cancellation rate calculation result data 106e in FIG. 49. Note that "2021/04" displayed on graph 3 is based on the accounting year and month held by the anomaly determination result data 106g.

The data binding information used when binding data to graph 4 and the generated graph 4 are shown in FIG. 55. The daily amount/number of times line graph generation unit 102i generates a line graph (graph 4 in FIG. 55) with the sales amount on the vertical axis and the registration processing date on the horizontal axis, and a line graph (graph 4 in FIG. 55) with the sales cancellation amount on the vertical axis and the registration processing date on the horizontal axis, for 2021/04 and department A, based on the daily sales cancellation data 106f in FIG. 51. Note that "2021/04" and "Department A" displayed in graph 4 are based on the accounting year/month and department, respectively, held by the anomaly determination result data 106g.

5. The generated graphs 1 to 4 are displayed in a list on the analysis screen, as shown in Figure 56.

6. Data for which an abnormality has been detected may be highlighted (e.g., displayed in a different color) based on department information contained in the abnormality determination result data 106g linked to the abnormality determination result message detail data 106i. For example, when displaying graph 2, as shown in FIG. 57, the color of the abnormality flag in the legend may be set to True (red) for the department to which the person in charge for whom the abnormality was detected belongs (= the department contained in the abnormality determination result data 106g), while the color of the abnormality flag in the legend may be set to False (gray) for departments other than the department in which the abnormality was detected.

In addition to graph 2, it is also possible to highlight (for example, by changing the color) using a similar method. The information related to sales cancellations is highlighted. That is, in the case of graph 1, the bar graph of "Sales Cancellation Amount" is highlighted. In the case of graph 3, the bar graph of "Negative Correction Amount," "Deletion Amount," and "Negative Sales Amount" is highlighted. And in the case of graph 4, the line graph of "Sales Cancellation Amount" is highlighted.

7. Here, the key point with graphs 1 to 4 displayed is to check, for example, whether a particular department has experienced a large amount of sales cancellations. If a particular department has experienced a large amount of sales cancellations, this means that there is a possibility that fraud is occurring on a departmental level. Furthermore, in the department where fraud is occurring, there is a possibility that multiple employees are performing the cancellation operations.

As a concrete example, Figure 58 shows a side-by-side comparison of Graph 2 (shown in Figure 32) when tabulated by person in charge and Graph 2 (shown in Figure 53) when tabulated by department. As Figure 58 shows, it can be seen that the fraud is committed in department A, and that within department A, person in charge A is also committing the fraud.

[5. Summary of the present embodiment]
In this way, according to the sales fraud detection operation support device 100 of this embodiment, as mainly described in [4-1], it is possible to support the operation of detecting sales fraud by automatically calculating the sales cancellation rate and detecting abnormally increasing values within the calculated sales cancellation rate.

In addition, according to the sales fraud detection business support device 100 of this embodiment, as mainly explained in [4-2], by referring to the four graphs displayed on the analysis screen, it is also possible to analyze the cause of the abnormality in the sales cancellation rate.

Conventionally, for sales transactions that occur on a regular basis, a person in charge could manually check the history at the time of data entry and detect abnormalities related to sales cancellations. However, this method requires specialized knowledge to collect various data and perform calculations before a sales fraud is discovered, which results in a lot of work time and a high risk of operational errors and missed checks.

In this embodiment, for example, data registered due to sales fraud can be automatically detected and notified within transaction data in sales operations. Specifically, the rate at which sales cancellations occur relative to sales in a single month can be calculated to detect abnormalities. In addition, a screen can be output that allows analysis of the timing, amount, and person in charge of processing sales and sales cancellations.

In other words, in this embodiment, for example, sales data is divided into data related to sales reductions and data related to sales increases, and each data is aggregated, making it possible to visualize the status of sales cancellations. In addition, the aggregation method and the alert method have also been devised.

As a result, this embodiment provides the following effects, for example. First, it is now possible to periodically and automatically detect sales fraud from a huge amount of sales transaction history information. In other words, even people without specialized knowledge can notice sales fraud. In addition, because the system automatically processes content that requires data to be compiled and compared under various conditions, the risk of operational errors can be avoided. And because a screen can be output that analyzes and confirms the abnormality and its basis, highly reliable detection of abnormalities is possible.

[6. Contribution to the United Nations-led Sustainable Development Goals (SDGs)]
This embodiment can contribute to improving business efficiency and promoting appropriate management decisions by companies, thereby making it possible to contribute to goals 8 and 9 of the SDGs.

In addition, this embodiment can contribute to reducing waste and promoting paperless and electronic systems, which can contribute to the achievement of SDGs goals 12, 13, and 15.

In addition, this embodiment can contribute to strengthening control and governance, making it possible to contribute to Goal 16 of the SDGs.

7. Other embodiments
The present invention may be embodied in various different embodiments other than those described above within the scope of the technical concept set forth in the claims.

For example, among the processes described in the embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using a known method.

In addition, the processing procedures, control procedures, specific names, registered data for each process, information including parameters such as search conditions, screen examples, and database configurations shown in this specification and drawings may be changed as desired unless otherwise specified.

In addition, with regard to the sales fraud detection business support device 100, each component shown in the figure is a functional concept, and does not necessarily have to be physically configured as shown in the figure.

For example, the processing functions of the sales fraud detection business support device 100, particularly the processing functions performed by the control unit, may be realized in whole or in part by a CPU and a program interpreted and executed by the CPU, or may be realized as hardware using wired logic. The program is recorded on a non-transient computer-readable recording medium that contains programmed instructions for causing the information processing device to execute the processes described in this embodiment, and is mechanically read by the sales fraud detection business support device 100 as necessary. That is, a computer program is recorded in a storage unit such as a ROM or HDD (Hard Disk Drive) to give instructions to the CPU in cooperation with the OS and perform various processes. This computer program is executed by being loaded into RAM, and cooperates with the CPU to form the control unit.

This computer program may also be stored in an application program server connected to the sales fraud detection business support device 100 via any network, and it may be possible to download all or part of it as needed.

In addition, the program for executing the processing described in this embodiment may be stored on a non-transitory computer-readable recording medium, and may also be configured as a program product. Here, the term "recording medium" refers to a memory card, a USB (Universal Serial Bus) memory, a SD (Secure Digital) card, a flexible disk, a magneto-optical disk, a ROM, an EPROM (Erasable Programmable Read Only Memory), an EEPROM (registered trademark) (Electrically Erasable and Programmable Read Only Memory), a CD-ROM (Compact Disk Read Only Memory), an MO (Magneto-Optical disk), a DVD (Digital Versatile This includes any "portable physical media" such as a Blu-ray (registered trademark) Disc, and a Blu-ray (registered trademark) Disc.

A "program" is a data processing method written in any language or description method, and may be in any format, such as source code or binary code. Note that a "program" is not necessarily limited to a single configuration, but also includes a distributed configuration consisting of multiple modules or libraries, and a program that works in conjunction with a separate program, such as an OS, to achieve its function. Note that the specific configuration and reading procedure for reading a recording medium in each device shown in the embodiments, as well as the installation procedure after reading, can use well-known configurations and procedures.

The various databases stored in the memory unit are storage devices such as RAM, ROM, etc., fixed disk devices such as hard disks, flexible disks, and optical disks, and store various programs, tables, databases, and web page files used for various processes and providing websites.

The sales fraud detection business support device 100 may be configured as an information processing device such as a known personal computer or workstation, or may be configured as the information processing device to which any peripheral device is connected. The sales fraud detection business support device 100 may be realized by implementing software (including programs or data, etc.) that causes the device to realize the processing described in this embodiment.

Furthermore, the specific form of distribution and integration of the devices is not limited to that shown in the figures, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various additions or functional loads. In other words, the above-mentioned embodiments can be implemented in any combination, or the embodiments can be implemented selectively.

The present invention is useful, for example, in any industry or business where sales performance management is required.

100 Sales fraud detection business support device 102 Control unit
102a Amount tallying section
102b Number of times counting unit
102c Sales cancellation rate calculation unit
102d Abnormal value detection unit
102e Abnormality display section
102f Period-Amount/Number of Times Bar Graph Generation Unit
102g Period-sales cancellation rate line graph generation section
102h Counting unit - amount and number of times bar graph generation section
102i Daily amount/number of times line graph generating unit 104 Communication interface unit 106 Storage unit
106a Automatic detection execution schedule data
106b Acquisition range condition data
106c Sales record history data
106d Sales amount and sales cancellation data
106e Sales cancellation rate calculation result data
106f Daily sales cancellation data
106g Abnormality determination result data
106h Abnormality judgment result message data
106i Abnormality determination result message detail data 108 Input/output interface unit 112 Input device 114 Output device 200 Server 300 Network

Claims (11)

A sales fraud detection operation support device that includes a control unit and a storage unit and is capable of supporting an operation for detecting fraud regarding sales results,
The storage unit includes:
The data includes a period, at least one of a unit of calculation from among establishment identification data, department identification data, and person in charge identification data, and at least one of a combination of a sales amount and a sales cancellation amount and a combination of a number of sales and a number of sales cancellations, and
The control unit is
a sales cancellation rate calculation means for calculating a sales cancellation rate for each combination of the period and the tabulation unit by dividing the sales cancellation amount in the sales amount/sales cancellation data by the sales amount in the sales amount/sales cancellation data when an amount is specified as the tabulation target, and by dividing the number of sales cancellations in the sales amount/sales cancellation data by the number of sales in the sales amount/sales cancellation data when a number is specified as the tabulation target;
an abnormal value detection means for detecting, according to a predetermined statistical method, a sales cancellation rate calculated by the sales cancellation rate calculation means that exceeds an upper limit as an abnormal value;
an abnormality display means for displaying a sales cancellation rate detected as an abnormal value by the abnormal value detection means and a combination of the period and the aggregation unit corresponding to the detected sales cancellation rate;
To have
A business support device for detecting sales fraud, characterized by: The storage unit includes:
Sales record history data including records having the period, the aggregation unit, and sales record amounts consisting of both positive and negative values is further stored;
The control unit is
further comprising an amount tallying means for tallying up the sales recorded amounts in the sales recording history data for each combination of the period and a specified tallying unit;
The amount tallying means is
calculates the sales cancellation amount for each combination of the period and the specified aggregation unit by aggregating the absolute value of the negative amount when the sales recognition amount in the newly registered record is a negative amount, the absolute value of the negative amount when the sum of the sales recognition amounts of the two records before and after the correction is a negative amount, and the absolute value of the negative sales recognition amount in the deleted record;
calculating the sales amount for each combination of the period and the specified aggregation unit by aggregating the recorded sales amounts in the records that were not included in the aggregation of the sales cancellation amounts;
2. The sales fraud detection business support device according to claim 1, The storage unit includes:
Sales record history data including records having the period, the aggregation unit, and sales record amounts consisting of both positive and negative values is further stored;
The control unit is
a counting means for counting the number of times the records in the sales recording history data are counted for each combination of the period and a specified counting unit;
The frequency counting means
if the sales amount in the newly registered record is a negative amount, then that one record is counted as one occurrence, if the sum of the sales amounts of two records before and after the correction is a negative amount, then those two records are counted as one occurrence, and one deleted record is counted as one occurrence, thereby calculating the number of sales cancellations for each combination of the period and the specified calculation unit;
calculating the number of sales for each combination of the period and the specified aggregation unit by counting the newly registered record as one transaction if the sales amount in the newly registered record is a positive amount, and counting the two records as one transaction if the sum of the sales amounts of the two records before and after the correction is a positive amount;
3. The sales fraud detection business support device according to claim 1 or 2. The abnormality display means is
Displaying the predetermined statistical method and the upper limit value together;
4. The sales fraud detection business support device according to claim 1, The predetermined statistical method is a method using an interquartile range,
the upper limit is calculated by adding 1.5 times the interquartile range to the third quartile;
5. The sales fraud detection business support device according to claim 1, The storage unit includes:
The sales cancellation rate calculation result data including the period, at least one of the tabulation units among the business establishment identification data, the department identification data, and the person in charge identification data, and at least one of the tabulation targets among the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations is further stored,
The control unit is
a period-amount/number-of-times bar graph generating means for generating a bar graph showing the sales amount and the sales cancellation amount for each period for the counting unit corresponding to the sales cancellation rate detected by the abnormal value detection means, or a bar graph showing the number of sales and the number of sales cancellations for each period for the counting unit corresponding to the sales cancellation rate detected by the abnormal value detection means, based on the sales cancellation rate calculation result data;
The bar graph is:
the vertical axis represents the sales amount and the sales cancellation amount in the sales cancellation rate calculation result data, or the number of sales and the number of sales cancellations in the sales cancellation rate calculation result data;
The time period in the sales cancellation rate calculation result data is plotted on the horizontal axis;
6. The sales fraud detection business support device according to claim 1, The storage unit includes:
further storing sales cancellation rate calculation result data including the period, at least one of the establishment identification data, the department identification data, and the person in charge identification data, and the sales cancellation rate calculated by the sales cancellation rate calculation means;
The control unit is
a period-sales cancellation rate line graph generating means for generating a line graph showing a change in the sales cancellation rate for each period for the number of types included in the specific aggregation unit based on the sales cancellation rate calculation result data,
The line graph is:
The sales cancellation rate in the sales cancellation rate calculation result data is set as the vertical axis,
The time period in the sales cancellation rate calculation result data is plotted on the horizontal axis;
7. The sales fraud detection business support device according to claim 1, The storage unit includes:
The sales cancellation rate calculation result data including the period, at least one of the tabulation units among the business establishment identification data, the department identification data, and the person in charge identification data, and at least one of the tabulation targets among the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations is further stored,
The control unit is
a bar graph generating means for generating a bar graph showing the sales amount and the sales cancellation amount for each of the counting units for the period corresponding to the sales cancellation rate detected by the abnormal value detection means, or a bar graph showing the number of sales and the number of sales cancellations for each of the counting units for the period corresponding to the sales cancellation rate detected by the abnormal value detection means, based on the sales cancellation rate calculation result data;
The bar graph is:
the vertical axis represents the sales amount and the sales cancellation amount in the sales cancellation rate calculation result data, or the number of sales and the number of sales cancellations in the sales cancellation rate calculation result data;
The aggregation unit in the sales cancellation rate calculation result data is set as a horizontal axis;
8. The sales fraud detection business support device according to claim 1, The storage unit includes:
daily sales cancellation data including the period, the registration processing date, at least one of the tabulation units of the business establishment identification data, the department identification data, and the person in charge identification data, and at least one of the tabulation targets of the combination of the sales amount and the sales cancellation amount and the combination of the number of sales and the number of sales cancellations is further stored;
The control unit is
a daily amount/number of times line graph generating means for generating, based on the daily sales cancellation data, a combination of two graphs, a line graph showing the sales amount for each registration processing date for the combination of the period and the tallying unit corresponding to the sales cancellation rate detected by the abnormal value detection means, and a line graph showing the sales cancellation amount for each registration processing date for the combination of the period and the tallying unit corresponding to the sales cancellation rate detected by the abnormal value detection means, or a combination of two graphs, a line graph showing the number of sales for each registration processing date for the combination of the period and the tallying unit corresponding to the sales cancellation rate detected by the abnormal value detection means, and a line graph showing the number of sales cancellations for each registration processing date for the combination of the period and the tallying unit corresponding to the sales cancellation rate detected by the abnormal value detection means,
The line graph is:
The vertical axis represents the sales amount, the sales cancellation amount, the number of sales, or the number of sales cancellations in the daily sales cancellation data,
The registration processing date in the daily sales cancellation data is plotted on the horizontal axis;
9. The sales fraud detection business support device according to claim 1, A sales fraud detection operation support method that can support an operation of detecting fraud regarding sales results, the method being executed by an information processing device having a control unit and a storage unit,
The storage unit includes:
The data includes a period, at least one of a unit of calculation from among establishment identification data, department identification data, and person in charge identification data, and at least one of a combination of a sales amount and a sales cancellation amount and a combination of a number of sales and a number of sales cancellations, and
Executed by the control unit,
a sales cancellation rate calculation step for calculating a sales cancellation rate for each combination of the period and the aggregation unit by dividing the sales cancellation amount in the sales amount/sales cancellation data by the sales amount in the sales amount/sales cancellation data when an amount is specified as the aggregation target, and by dividing the number of sales cancellations in the sales amount/sales cancellation data by the number of sales in the sales amount/sales cancellation data when a number is specified as the aggregation target;
an abnormal value detection step of detecting, according to a predetermined statistical method, a sales cancellation rate that exceeds an upper limit value among the sales cancellation rates calculated in the sales cancellation rate calculation step, as an abnormal value;
an abnormality display step of displaying a sales cancellation rate detected as an abnormal value in the abnormal value detection step and a combination of the period and the aggregation unit corresponding to the detected sales cancellation rate;
containing,
A method for supporting sales fraud detection operations, comprising: A sales fraud detection support program for supporting a task of detecting fraud relating to sales results, the task being executed by an information processing device having a control unit and a storage unit, the task comprising:
The storage unit includes:
The data includes a period, at least one of a unit of calculation from among establishment identification data, department identification data, and person in charge identification data, and at least one of a combination of a sales amount and a sales cancellation amount and a combination of a number of sales and a number of sales cancellations, and
To cause the control unit to execute
a sales cancellation rate calculation step for calculating a sales cancellation rate for each combination of the period and the aggregation unit by dividing the sales cancellation amount in the sales amount/sales cancellation data by the sales amount in the sales amount/sales cancellation data when an amount is specified as the aggregation target, and by dividing the number of sales cancellations in the sales amount/sales cancellation data by the number of sales in the sales amount/sales cancellation data when a number is specified as the aggregation target;
an abnormal value detection step of detecting, according to a predetermined statistical method, a sales cancellation rate that exceeds an upper limit value among the sales cancellation rates calculated in the sales cancellation rate calculation step, as an abnormal value;
an abnormality display step of displaying a sales cancellation rate detected as an abnormal value in the abnormal value detection step and a combination of the period and the aggregation unit corresponding to the detected sales cancellation rate;
containing,
A program to assist in the detection of sales fraud.

Images